Retry transient API errors with exponential backoff

[adamlogic]

Transient network conditions can result in timeouts that can often be retried successfully.

---

Note

Medium Risk
Adds retry/backoff behavior around metrics posting; low complexity but changes runtime network behavior and could affect reporting latency under failure conditions.

Overview
Reporter now retries transient network failures when posting metrics. Reporter._report_metrics retries up to MAX_RETRIES (3) on ConnectionError/Timeout with exponential backoff, and logs a dedicated warning if it still cannot connect.

Non-transient requests errors (e.g., HTTP failures) still log once and abort without retry. Tests were added to assert retry vs non-retry behavior while patching requests.post/time.sleep.

^{Written by Cursor Bugbot for commit ab3f002. This will update automatically on new commits. Configure here.}

[carlosantoniodasilva]

👍 nice.